Abstract
M.A. (Industrial Psychology)
Two main aims were assessed in this study. Firstly, the intention was to investigate structural invariance in Holland’s circular/hexagonal model across gender (i.e., to investigate whether or not there is construct or structural equivalence across gender). The second aim was to evaluate if mean score differences in the RIASEC types (Realistic, Investigative, Artistic, Social, Enterprising, and Conventional) and the People-Things and Data-Ideas dimensions across gender exist. This study also investigated if item level bias (differential item functioning) across gender had an impact on the structure of Holland’s circular/hexagonal model or mean scores on the RIASEC types. In total 407 men and women who completed the South African Career Interest Inventory (SACII) participated in the study. A two-way analysis of variance was used to investigate uniform and non-uniform differential item functioning. Only 18 of the 142 SACII items were found to have differential item functioning.
The structure of Holland’s vocational/personality model was examined using multidimensional scaling, the randomisation test of hypothesised order relations, and circumplex covariance structure modelling. RIASEC ordering was detected within the obtained data, thus supporting international and South African research. Although the model fit both genders sufficiently and was mostly invariant across gender, the model produced a better fit for women when compared to men. The results are different from earlier South African research that noted that Holland’s circular/hexagonal model was not appropriate for the South African population. Although a better fit was observed for women, the results from this study propose that Holland’s circular ordering/hexagonal model might be more applicable in South Africa than what was suggested in earlier research. Differential item functioning had little impact on the structure of Holland’s circular/hexagonal model as the results held true regardless of whether or not DIF items were removed.
The study further investigated if mean score differences in the RIASEC types and the People-Things and Data-Ideas dimensions exist between men and women and if these differences could be attributed to items with differential item functioning. Mean score differences were investigated using independent samples t-tests and discriminant analysis. Mean score differences were noted on the Realistic scale (with men receiving a higher score) and the Social scale (with women scoring higher). Men also scored higher on the Things pole and women higher on the People pole. The mean score differences remained even after items with differential item